Model Context Protocol Marketing Automation
The Model Context Protocol gives AI agents a structured way to call external tools. marginal applies that to email: a hosted MCP server where your agent generates variants, launches tests, and reads results without leaving the conversation.
What Model Context Protocol marketing automation looks like
Traditional marketing automation lives in a dashboard. You log in, draft variants, set up an experiment, wait, then come back to read the numbers. Model Context Protocol marketing automation collapses that loop. Your agent speaks MCP, marginal exposes email tools over MCP, and the whole workflow happens through tool calls.
Because the protocol is standardized, the same marginal endpoint works across agents and IDEs. You describe the campaign in plain language; the agent picks the right tool and fills in the arguments.
- generate_variants — draft subject line candidates for a test
- launch_test — start a managed send and split the audience
- get_results — pull opens, clicks, and within-test lift
- recommend_next — get a data-backed suggestion for the follow-up send
Running a subject line A/B test from your agent
A subject line A/B test is the most common automation people reach for, and it maps cleanly to marginal's tools. Ask your agent for several subject variants, launch the test as a managed send, and let marginal handle the split and tracking.
Open and click tracking is built in, and lift is reported within the test itself, so you compare variants against each other rather than against stale historical baselines.
- Generate 3–5 variants, then trim to the strongest before launch
- marginal manages the send — no separate delivery setup to wire up
- Read open/click rates and lift once recipients engage
- Use recommend_next to carry the winner into your next campaign
Why a hosted email MCP server fits AI agent email marketing
marginal is a hosted email MCP server, so there is nothing to deploy or self-host. You point your client at https://marginal.sh/mcp, authenticate with a Bearer API key, and your agent can start running experiments.
For AI agent email marketing this matters: the agent gets a stable, always-on endpoint instead of brittle local tooling. The free tier covers 100 experiments per month, which is plenty for testing the workflow and shipping early campaigns.
- Endpoint: https://marginal.sh/mcp
- Auth: Bearer API key (marg_live_...)
- Works with Cursor, Claude Desktop, Claude Code, Windsurf, Cline, Continue, Codex, and Zed
- Free tier: 100 experiments/month
Getting started
Connect marginal to the agent or IDE you already use, then run a first subject line A/B test to see the loop end to end. The docs at https://marginal.sh/docs/ walk through each tool and the arguments it accepts.
From there, automation is mostly prompting: describe the audience and goal, and let your agent string generate_variants, launch_test, get_results, and recommend_next together into a repeatable optimization cycle.
- Add the https://marginal.sh/mcp endpoint to your MCP client
- Run generate_variants for your first campaign idea
- Launch a test and check results once engagement lands
- Read the full tool reference at https://marginal.sh/docs/
Get started with marginal
marginal is a hosted email marketing MCP server at marginal.sh. Sign up free, create an API key, and connect https://marginal.sh/mcp from Cursor, Codex, or Claude Desktop.
- 100 experiments/month on the free tier
- Four MCP tools: generate_variants, launch_test, get_results, recommend_next
- Listed in the MCP Registry as sh.marginal/mcp